The Neoverse N1 platform is optimized for 7nm process technology and focuses on compute speed.
“This N1 platform is really about the core compute you need in the hyperscale data center, at the 5G base station, or at an internet gateway,” said Drew Henry, Arm’s SVP and GM for infrastructure.
This chip design performs better than previous Arm CPU generations: the company claims it can deliver 2.5 times more performance on cloud workloads such as Nginx and MemcacheD. It can also scale from 4 to 128 cores, which Arm says gives its chipmaker partners the ability to build more diverse products by adding accelerators or other features with their own on-chip custom silicon.
“It also has a series of features built in: virtualization technologies, performance and management technologies, reliability and service support, the ability to do profiling as applications are being executed,” Henry said.
The second platform, Neoverse E1, focuses more on throughput performance. It is designed to help companies transition from 4G to 5G infrastructure, and Arm says it achieves 2.7 times more throughput performance, 2.4 times more throughput efficiency, and over 2 times more compute performance compared to previous generations.
Taking On Intel
Arm doesn’t build chips. Instead it provides intellectual property (IP), which it licenses to semiconductor companies that embed it into their chips. Although chips for smart phones and other mobile devices have been ARM’s most successful business, the company has been branching out in recent years, developing IP for IoT devices, cars, laptops, and data center servers.
Today’s announcement continues that push — and sets up Arm, albeit indirectly, to further compete against Intel in the lucrative data center server market.
Last month Huawei announced a new CPU, called Kunpeng 920, based on Arm — not Intel — architecture. And in December semiconductor startup Ampere Computing released its first Arm-based data center chip and said it is on track to roll out its 7nm chip also based on Arm architecture this year.
Drew said Arm partners will release processors based on N1 and E1 this year.
“Heterogeneous computing is really where the industry is moving,” Drew said. “In this post-Moore’s Law world you need to be able to provide a very flexible design that allows our customers to really build what they want to build, and that’s why we are seeing such remarkable things happening in the silicon ecosystem.”